云辅助的安全高效非负矩阵分解算法

A secure and efficient cloud-assisted algorithm for non-negative matrix factorization

为了改善非负矩阵分解(Non-negative Matrix Factorization,NMF)进行大规模数据降维时消耗计算资源的问题,提出一种基于单服务器的安全外包算法。首先对输入矩阵使用随机矩阵填充,然后对填充后的矩阵使用随机对角矩阵变换和随机置换进行加密,盲化原始矩阵中非零项及零元素的数目和分布。理论分析和实验结果均表明,与不外包情形相比,所提算法能使本地端获得可观的计算节省。

To address the issue of computing resource consumption in Non-negative Matrix Factorization(NMF)for large-scale data dimensionality reduction,a secure outsourcing algorithm based on single server is proposed.Firstly,the input matrix is filled with a random matrix,and then the filled matrix is encrypted by random diagonal matrix transformation and random permutation to blind the number and distribution of nonzero entries and zero elements in the original matrix.Both theoretical analysis and experimental results show that,comparing with the scenario without outsourcing,the proposed algorithm significantly reduces the computational burden on the local end,leading to substantial savings in computation resources.